A Novel Driving Noise Analysis Method for On-Road Traffic Detection.

Area of Science:

• Signal Processing
• Acoustics
• Transportation Engineering

Background:

• Effective noise reduction and abnormal feature extraction are crucial for abnormal sound detection in urban traffic.
• Traditional methods like Short-Time Energy (STE) and Mel Frequency Cepstral Coefficients (MFCC) have limitations in detecting continuous traffic flow and overlapping vehicles.

Purpose of the Study:

• To explore effective methods for improving traffic noise detection accuracy.
• To address the limitations of traditional methods in handling continuous traffic flow and overlapping vehicle sounds.
• To propose an innovative traffic noise detection solution.

Main Methods:

• An improved spectral subtraction method was employed for traffic noise analysis.
• Feature fusion of STE and MFCC coefficients created an enhanced feature parameter, E-MFCC.
• A novel traffic noise detection solution based on Triangular Wave Analysis (TWA) was developed.
• A traffic detection simulation platform was established using MATLAB's APP Designer.

Main Results:

• The proposed Triangular Wave Analysis (TWA) achieved a detection accuracy of 91%.
• This significantly surpasses the accuracies of traditional STE (67.77%) and MFCC (76.01%) methods.
• The method effectively addresses the challenge of detecting overlapping vehicle sounds.

Conclusions:

• The proposed TWA method demonstrates superior performance in abnormal sound detection for urban traffic.
• This approach enhances the accuracy of road traffic volume detection, particularly in complex scenarios.
• The findings contribute to improving the efficiency of road traffic operations through advanced acoustic analysis.